Vacancy-screw dislocation interactions in bcc Nb: A machine-learning potential investigation

Vacancy-screw dislocation interactions in bcc Nb are studied with near density functional theory accuracy using a machine-learning potential, specifically a moment tensor potential (MTP), in large-scale atomistic simulations. The MTP predicts attractive vacancy-dislocation interactions with threefold symmetry in the (111) plane. These attractive interactions lead to a decrease in the Peierls stress by up to 28%, depending on the vacancy position relative to the dislocation. The interactions also lead to the formation of new complex topological defects (CTDs), when the vacancy is close to the dislocation core. A CTD can be geometrically characterized as a screw dislocation with and with a (partially) vacancy along the dislocation line. CTDs form under shear when the vacancy-dislocation distance is less than ∼6 Å , providing a new mechanism for pinning screw dislocations. The formation of CTDs confirms theoretical predictions made in the early 1990s based on differential geometry and underscores the importance of machine-learning potentials and large-scale simulations for accurately describing the properties of extended defects in metals.